[Bug c++/40834] New: [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[Bug c++/40834] New: [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

+++ This bug was initially created as a clone of Bug #40799 +++

With revision 149946, on Linux/x86-64, the problem now became

  Running 483.xalancbmk ref base o3 default

483.xalancbmk: copy 0 non-zero return code (exit code=0, signal=11)

It happens with both "-O3 -ffast-math" and "-O2 -ffast-math".


-- 
           Summary: [4.5 Regression] Revision 149750 failed 483.xalancbmk in
                    SPEC CPU 2006
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: hjl dot tools at gmail dot com
 BugsThisDependsOn: 40799


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[rguenth at gcc dot gnu dot org]

------- Comment #1 from rguenth at gcc dot gnu dot org  2009-07-23 08:47 -------
Can you try with the patch for 40799 applied before gimplification
unit-at-a-time? Does the failure reproduce with the test data or only with the
ref data?


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #2 from hjl dot tools at gmail dot com  2009-07-23 14:40 -------
(In reply to comment #1)
> Can you try with the patch for 40799 applied before gimplification
> unit-at-a-time? Does the failure reproduce with the test data or only with the
> ref data?
> 

I applied the patch for PR 40799 on revision 149748 and
I got the same failure with both test/ref input data.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[rguenther at suse dot de]

------- Comment #3 from rguenther at suse dot de  2009-07-23 15:21 -------
Subject: Re:  [4.5 Regression] Revision 149750 failed
 483.xalancbmk in SPEC CPU 2006

On Thu, 23 Jul 2009, hjl dot tools at gmail dot com wrote:

> ------- Comment #2 from hjl dot tools at gmail dot com  2009-07-23 14:40 -------
> (In reply to comment #1)
> > Can you try with the patch for 40799 applied before gimplification
> > unit-at-a-time? Does the failure reproduce with the test data or only with the
> > ref data?
> > 
> 
> I applied the patch for PR 40799 on revision 149748 and
> I got the same failure with both test/ref input data.

Ok, I will have a look next week.  If you can figure out which source
file of 483.xalancbmk has the relevant change until then you receive
extra bonus points ;)  I'm really curious that dropping lang-specific
tree earlier causes problems ...

Richard.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #4 from hjl dot tools at gmail dot com  2009-07-23 18:15 -------
FunctionSubstringAfter.cpp  FunctionSubstringBefore.cpp  FunctionSubstring.cpp
are miscompiled. I have to replace all of them to get a working binary,


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #5 from hjl dot tools at gmail dot com  2009-07-23 18:19 -------
(In reply to comment #4)
> FunctionSubstringAfter.cpp  FunctionSubstringBefore.cpp  FunctionSubstring.cpp
> are miscompiled. I have to replace all of them to get a working binary,
> 

It fails even with -O0. Replace those 3 files gives me a working binary.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #6 from hjl dot tools at gmail dot com  2009-07-23 18:48 -------
Created an attachment (id=18246)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18246&action=view)
A testcase

You can compare the outputs before and after change.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #7 from hjl dot tools at gmail dot com  2009-07-23 18:56 -------
Here are the differences between good and bad

GOOD const struct XalanDOMString & theFirstString = (const struct
XalanDOMString &) (const struct XalanDOMString *) OBJ_TYPE_REF(*(SAVE_EXPR
<xalanc_1_8::XObjectPtr::operator->
(&arg1)>->D.20005._vptr.XalanReferenceCountedObject + 72);SAVE_EXPR
<xalanc_1_8::XObjectPtr::operator-> (&arg1)>->9) (SAVE_EXPR
<xalanc_1_8::XObjectPtr::operator-> (&arg1)>);
BAD const struct XalanDOMString & theFirstString = (const struct XalanDOMString
&) (const struct XalanDOMString *) OBJ_TYPE_REF(*(SAVE_EXPR
<xalanc_1_8::XObjectPtr::operator-> ((const struct XObjectPtr *)
arg1)>->D.20005._vptr.XalanReferenceCountedObject + 72);SAVE_EXPR
<xalanc_1_8::XObjectPtr::operator-> ((const struct XObjectPtr *) arg1)>->9)
(SAVE_EXPR <xalanc_1_8::XObjectPtr::operator-> ((const struct XObjectPtr *)
arg1)>);

That is

(&arg1)>->D.20005._vptr.XalanReferenceCountedObject + 72)

vs.

((const struct XObjectPtr *) arg1)>->D.20005._vptr.XalanReferenceCountedObject
+ 72)

The correct one is an address. The wrong one is a cast to pointer.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #8 from hjl dot tools at gmail dot com  2009-07-23 18:57 -------
Jason, can you take a look at this?


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at redhat dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #9 from hjl dot tools at gmail dot com  2009-07-23 19:16 -------
This patch:

Index: cp-gimplify.c
===================================================================
--- cp-gimplify.c       (revision 149933)
+++ cp-gimplify.c       (working copy)
@@ -804,15 +804,6 @@ cp_genericize_r (tree *stmt_p, int *walk
        }
     }

-  else if (TREE_CODE (stmt) == BIND_EXPR)
-    {
-      VEC_safe_push (tree, heap, wtd->bind_expr_stack, stmt);
-      cp_walk_tree (&BIND_EXPR_BODY (stmt),
-                   cp_genericize_r, data, NULL);
-      VEC_pop (tree, wtd->bind_expr_stack);
-      *walk_subtrees = 0;
-    }
-
   else if (TREE_CODE (stmt) == USING_STMT)
     {
       tree block = NULL_TREE;

fixed the problem. Does it make any senses?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[rguenther at suse dot de]

------- Comment #10 from rguenther at suse dot de  2009-07-24 09:02 -------
Subject: Re:  [4.5 Regression] Revision 149750 failed
 483.xalancbmk in SPEC CPU 2006

On Thu, 23 Jul 2009, hjl dot tools at gmail dot com wrote:

> ------- Comment #9 from hjl dot tools at gmail dot com  2009-07-23 19:16 -------
> This patch:
> 
> Index: cp-gimplify.c
> ===================================================================
> --- cp-gimplify.c       (revision 149933)
> +++ cp-gimplify.c       (working copy)
> @@ -804,15 +804,6 @@ cp_genericize_r (tree *stmt_p, int *walk
>         }
>      }
> 
> -  else if (TREE_CODE (stmt) == BIND_EXPR)
> -    {
> -      VEC_safe_push (tree, heap, wtd->bind_expr_stack, stmt);
> -      cp_walk_tree (&BIND_EXPR_BODY (stmt),
> -                   cp_genericize_r, data, NULL);
> -      VEC_pop (tree, wtd->bind_expr_stack);
> -      *walk_subtrees = 0;
> -    }
> -
>    else if (TREE_CODE (stmt) == USING_STMT)
>      {
>        tree block = NULL_TREE;
> 
> fixed the problem. Does it make any senses?

Not really.  It will break debug information for using declarations.

If you add cp_walk_trees to BIND_EXPR_VARS and/or BIND_EXPR_BLOCK
does that fix it?

Richard.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #11 from hjl dot tools at gmail dot com  2009-07-24 14:16 -------
(In reply to comment #10)
> Subject: Re:  [4.5 Regression] Revision 149750 failed
>  483.xalancbmk in SPEC CPU 2006
> 
> On Thu, 23 Jul 2009, hjl dot tools at gmail dot com wrote:
> 
> > ------- Comment #9 from hjl dot tools at gmail dot com  2009-07-23 19:16 -------
> > This patch:
> > 
> > Index: cp-gimplify.c
> > ===================================================================
> > --- cp-gimplify.c       (revision 149933)
> > +++ cp-gimplify.c       (working copy)
> > @@ -804,15 +804,6 @@ cp_genericize_r (tree *stmt_p, int *walk
> >         }
> >      }
> > 
> > -  else if (TREE_CODE (stmt) == BIND_EXPR)
> > -    {
> > -      VEC_safe_push (tree, heap, wtd->bind_expr_stack, stmt);
> > -      cp_walk_tree (&BIND_EXPR_BODY (stmt),
> > -                   cp_genericize_r, data, NULL);
> > -      VEC_pop (tree, wtd->bind_expr_stack);
> > -      *walk_subtrees = 0;
> > -    }
> > -
> >    else if (TREE_CODE (stmt) == USING_STMT)
> >      {
> >        tree block = NULL_TREE;
> > 
> > fixed the problem. Does it make any senses?
> 
> Not really.  It will break debug information for using declarations.
> 
> If you add cp_walk_trees to BIND_EXPR_VARS and/or BIND_EXPR_BLOCK
> does that fix it?
> 
> Richard.
> 
I tried

Index: cp-gimplify.c
===================================================================
--- cp-gimplify.c       (revision 149933)
+++ cp-gimplify.c       (working copy)
@@ -807,6 +807,10 @@ cp_genericize_r (tree *stmt_p, int *walk
   else if (TREE_CODE (stmt) == BIND_EXPR)
     {
       VEC_safe_push (tree, heap, wtd->bind_expr_stack, stmt);
+      cp_walk_tree (&BIND_EXPR_VARS (stmt),
+                   cp_genericize_r, data, NULL);
+      cp_walk_tree (&BIND_EXPR_BLOCK (stmt),
+                   cp_genericize_r, data, NULL);
       cp_walk_tree (&BIND_EXPR_BODY (stmt),
                    cp_genericize_r, data, NULL);
       VEC_pop (tree, wtd->bind_expr_stack);

It doesn't help. 


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[rguenth at gcc dot gnu dot org]

------- Comment #12 from rguenth at gcc dot gnu dot org  2009-07-24 19:34 -------
Hm, I'm quite lost here.  But at least here is a simplified testcase which
shows the same differences.

class XalanDOMString { };
class XObject {
public:
    virtual const XalanDOMString&  str() const;
};
class XObjectPtr {
public:
    XObjectPtr(const XObjectPtr&);
    const XObject* operator->() const;
};
class FunctionSubstringBefore
{
  void execute(const XObjectPtr arg1) const;
};
void
FunctionSubstringBefore::execute(const XObjectPtr arg1) const
{
  const XalanDOMString& theFirstString = arg1->str();
}


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-07-24 19:34:08
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[rguenth at gcc dot gnu dot org]

------- Comment #13 from rguenth at gcc dot gnu dot org  2009-07-24 19:52 -------
I see what goes wrong.  preparing a patch.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rguenth at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2009-07-24 19:34:08         |2009-07-24 19:52:12
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[rguenth at gcc dot gnu dot org]

------- Comment #14 from rguenth at gcc dot gnu dot org  2009-07-24 19:59 -------
Index: cp/cp-gimplify.c
===================================================================
--- cp/cp-gimplify.c    (revision 150041)
+++ cp/cp-gimplify.c    (working copy)
@@ -804,10 +804,14 @@ cp_genericize_r (tree *stmt_p, int *walk
        }
     }

-  else if (TREE_CODE (stmt) == BIND_EXPR)
+  else if (TREE_CODE (stmt) == BIND_EXPR
+          /* We walk this twice, once to keep track of the BIND_EXPR
+             stack and once to properly walk the BIND_EXPR vars.  */
+          && (VEC_empty (tree, wtd->bind_expr_stack)
+              || VEC_last (tree, wtd->bind_expr_stack) != stmt))
     {
       VEC_safe_push (tree, heap, wtd->bind_expr_stack, stmt);
-      cp_walk_tree (&BIND_EXPR_BODY (stmt),
+      cp_walk_tree (stmt_p,
                    cp_genericize_r, data, NULL);
       VEC_pop (tree, wtd->bind_expr_stack);
       *walk_subtrees = 0;


in case you want to do some testing.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[rguenth at gcc dot gnu dot org]

------- Comment #15 from rguenth at gcc dot gnu dot org  2009-07-24 20:01 -------
The short explanation is that we did not properly walk DECL_INITIAL of the
vars in the BIND_EXPR.  Not with walking BIND_EXPR_VARS either because we
stop at VAR_DECLs in cp_genericize_r before walking its DECL_INITIAL.
walk_tree does not have this issue because it does not walk the VAR_DECLs
directly but only certain fields of it.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #16 from hjl dot tools at gmail dot com  2009-07-24 20:06 -------
(In reply to comment #14)
> Index: cp/cp-gimplify.c
> ===================================================================
> --- cp/cp-gimplify.c    (revision 150041)
> +++ cp/cp-gimplify.c    (working copy)
> @@ -804,10 +804,14 @@ cp_genericize_r (tree *stmt_p, int *walk
>         }
>      }
> 
> -  else if (TREE_CODE (stmt) == BIND_EXPR)
> +  else if (TREE_CODE (stmt) == BIND_EXPR
> +          /* We walk this twice, once to keep track of the BIND_EXPR
> +             stack and once to properly walk the BIND_EXPR vars.  */
> +          && (VEC_empty (tree, wtd->bind_expr_stack)
> +              || VEC_last (tree, wtd->bind_expr_stack) != stmt))
>      {
>        VEC_safe_push (tree, heap, wtd->bind_expr_stack, stmt);
> -      cp_walk_tree (&BIND_EXPR_BODY (stmt),
> +      cp_walk_tree (stmt_p,
>                     cp_genericize_r, data, NULL);
>        VEC_pop (tree, wtd->bind_expr_stack);
>        *walk_subtrees = 0;
> 
> 
> in case you want to do some testing.
> 

It works.  Thanks.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl dot tools at gmail dot com]

------- Comment #17 from hjl dot tools at gmail dot com  2009-07-24 21:55 -------
Created an attachment (id=18251)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18251&action=view)
A testcase

[hjl@gnu-34 bad]$ make
/export/gnu/import/rrs/149748-fixed/usr/bin/g++ -S pr40834.C -o bad.s
/export/gnu/import/rrs/149748/usr/bin/g++ -S pr40834.C -o good.s
g++ -o bad bad.s
g++ -o good good.s
./good
./bad
make: *** [all] Aborted


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #18246|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[rguenth at gcc dot gnu dot org]

------- Comment #19 from rguenth at gcc dot gnu dot org  2009-07-29 20:16 -------
Fixed.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[rguenth at gcc dot gnu dot org]

------- Comment #18 from rguenth at gcc dot gnu dot org  2009-07-29 20:16 -------
Subject: Bug 40834

Author: rguenth
Date: Wed Jul 29 20:16:32 2009
New Revision: 150222

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=150222
Log:
2009-07-29  Richard Guenther  <rguenther@suse.de>

        PR c++/40834
        * cp-gimplify.c (cp_genericize_r): Properly walk the BIND_EXPR
        vars.

        * g++.dg/torture/pr40834.C: New testcase.

Added:
    trunk/gcc/testsuite/g++.dg/torture/pr40834.C
Modified:
    trunk/gcc/cp/ChangeLog
    trunk/gcc/cp/cp-gimplify.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834

[Bug c++/40834] [4.5 Regression] Revision 149750 failed 483.xalancbmk in SPEC CPU 2006

[hjl at gcc dot gnu dot org]

------- Comment #20 from hjl at gcc dot gnu dot org  2009-08-05 14:45 -------
Subject: Bug 40834

Author: hjl
Date: Wed Aug  5 14:45:15 2009
New Revision: 150487

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=150487
Log:
2009-07-28  H.J. Lu  <hongjiu.lu@intel.com>

        Backport from mainline:
        2009-07-30  Martin Jambor  <mjambor@suse.cz>

        PR tree-optimization/40570
        * gcc.c-torture/compile/pr40570.c: New test.

        2009-07-29  Richard Guenther  <rguenther@suse.de>

        PR c++/40834
        * g++.dg/torture/pr40834.C: New testcase.

Added:
    branches/gcc-4_4-branch/gcc/testsuite/g++.dg/torture/pr40834.C
      - copied unchanged from r150485,
trunk/gcc/testsuite/g++.dg/torture/pr40834.C
    branches/gcc-4_4-branch/gcc/testsuite/gcc.c-torture/compile/pr40570.c
      - copied unchanged from r150484,
trunk/gcc/testsuite/gcc.c-torture/compile/pr40570.c
Modified:
    branches/gcc-4_4-branch/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40834